Unique Predictors and Economic Burden of Superficial and Deep/Organ Space Surgical Site Infections following Pancreatectomy

Abstract

Background

Surgical site infections (SSIs) are common following pancreatectomy and associated with significant morbidity and economic burden. We sought to identify distinct predictors for the development of superficial versus deep/organ space SSIs and their effects on surgical outcomes.

Methods

ACS-NSQIP targeted pancreatectomy 2014 and 2015 databases were queried. Univariate and multivariate models were developed for both types of SSI, length of stay (LOS), and readmission. Costs were estimated based on the Centers for Medicare & Medicaid Services (CMS) recommendations.

Results

Of 8 093 patients, there were 422 (5.2%) superficial and 1 005 (12.4%) deep/organ space SSIs. On multivariate analyses, preoperative biliary stenting was a predictor only for superficial SSI (OR: 2.21), while BMI of 25–29.9 (OR: 1.25) and BMI ≥30kg/m² (OR: 1.53), pancreatic duct size <3mm (OR: 1.30), and intermediate (OR: 1.67) versus hard gland texture were predictors of deep/organ-space SSI. Superficial and deep/organ space SSIs were independent predictors of prolonged LOS (OR 1.74 vs 1.80) and readmission (OR 2.59 vs 6.57). Additional readmission costs per patient secondary to superficial SSI and deep/organ space SSI were $7 661.37 and $18 409.42, respectively.

Conclusion

Deep/organ space SSI contributes more profoundly to prolonged hospital stay, readmission, and additional costs, suggesting that strategies should focus on preferential prevention of deep/organ space infections.

BACKGROUND

Surgical site infection (SSI) is a common complication following abdominal surgery. Rates of SSIs as high as 35% after pancreatectomy have been reported,^{1, 2} which is significantly higher than other major general surgical procedures, such as hepatectomy (3.1% to 14 %),³ and colectomy (5% to 26%).⁴ Differences in SSIs have been established, resulting in the classification by both the United States Centers for Disease Control and Prevention (CDC) ⁵ and the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) ⁶ into three distinct types, based on the level of tissue involvement: superficial incisional, deep incisional, and organ-space.

Studies in other abdominal surgical specialties have shown variations in risk factors among different SSI categories.^{3, 4, 7, 8, 9} In addition, surgical outcomes and treatment modalities differ depending on the SSI type, with deep or organ space SSIs often requiring more intensive and invasive interventions compared with superficial SSI. Despite these differences, many studies in pancreatic surgery investigating the risk factors and effects of SSI on patient outcomes have over time grouped all SSI’s together.^{1, 2, 10, 11, 12, 13} This may eliminate the opportunity to identify unique modifiable risk factors for proper targeting of quality improvement programs and cost reduction.

This study aims to detect unique predictors for the development of superficial and deep/organ space infections and evaluate their effects on readmission and length of hospital stay, as well as potential cost implications. Our findings hope to provide more granular data and direction for future quality improvement initiatives.

METHODS

Data Source and Study Population

A retrospective review of the ACS-NSQIP 2014 and 2015 targeted pancreatectomy participant use data file (PUF) was performed for all adult (age ≥18 years) patients.⁶ The ACS-NSQIP is a nationally validated, risk adjusted, peer-controlled registry of patient risk factors and 30-day postoperative outcomes aimed to improve quality of surgical care.^{14, 15} It is also de-identified and Health Insurance Portability and Accountability Act of 1996 (HIPAA) compliant.^{16, 17} Inclusion and exclusion criteria, sampling algorithms, and the collected variables with their definitions are publicly available through the American College of Surgeons web page.⁶ Institutional Review Board approval was obtained from our institution.

Patient Selection

Pancreatic procedures were identified by CPT codes (48120, 48140, 48145, 48146, 48148, 48150, 48152, 48153, 48154, 48155 and 48999). Only inpatients with electively planned pancreatectomies were included. The following groups of patients were excluded from the study: outpatients, disseminated cancer, emergency or non-elective procedures (as defined by the NSQIP manual). Patients with preoperative sepsis within 30 days of surgery, sepsis present at time of surgery (PATOS), septic shock PATOS, superficial SSI PATOS, deep/organ space SSI PATOS, pneumonia PATOS, and urinary tract infection PATOS were excluded (n = 1 924) since they may not be considered elective procedures due to another potential underlying etiology. Patients with missing values (height, weight, age, operation time, LOS, unplanned readmission) and unknown values (race, ASA class, radiotherapy, chemotherapy, fistula, vascular resection, operative approach, reason for unplanned readmission) and operation time of zero minutes were also excluded (n =1 202).

Patient Characteristics and Variables

The baseline, intra-operative and post-operative characteristics of all patients were analyzed. Albumin was categorized according to clinical significance: ≤ 3mg/dl or >3mg/dl and unknown albumin levels were treated as missing values. All patients who were 90 years and older were coded as a maximum of 90 years in NSQIP. Pancreatectomy type was grouped, based on current procedural terminology (CPT) codes as reported in NSQIP, into 1) proximal (48154, 48150, 48153, 48152), 2) distal (48146, 48140, 48145) and 3) others (48155, 48148, 48120, 48999). Days from operation until development of deep incisional SSI complication were used where days from operation until organ-space SSI were missing or unknown. For univariate analysis, age and operative duration were categorized as greater than the median or less than or equal to the median, and modeled as 10-year and 1-hour intervals, respectively, for multivariate regression analyses.

Outcomes

Primary outcomes were superficial SSI and deep and/or organ-space SSI within 30-days following pancreatectomy. NSQIP defines superficial SSI as infections involving the skin or subcutaneous tissue of the incision. Deep SSI is defined as an infection involving the deep soft tissues (e.g. fascial and muscle layers) of the incision, while organ-space SSI includes infections involving any part of the anatomy that was opened or manipulated during an operation (other than the incision). The detailed inclusion and exclusion criteria for each SSI are available in the ACS-NSQIP operations manual.⁶ Deep and organ-space SSI were grouped together, since infections originating from the organ-space and draining through the surgical incision are labeled as deep SSI in NSQIP, which is also consistent with the CDC definition.^{5, 6} Patients who developed both a superficial and a deep/organ space SSI were classified as deep/organ-space SSI since it supersedes a superficial SSI in most cases and is consistent with CDC definitions. ⁵ For analyses, the patient cohort was further divided into two groups: 1) patients with superficial SSI versus no SSI and 2) patients with deep/organ-space versus no SSI. Secondary outcomes of interest were: 1) length of hospital stay (LOS) measured as days from operation until discharge (modeled as a binary variable; greater than versus less than or equal to the median LOS)¹⁸ and 2) unplanned readmission within 30 days.

Medicare Costs

Costs were defined based on recommendations by the Centers for Medicare & Medicaid Services (CMS). ^{19, 20, 21, 22, 23} Readmission costs secondary to SSIs were calculated using Medicare’s relative value units (RVUs), hospital and physician fees which are made publicly available on the Centers for Medicare & Medicaid Services (CMS) website.^{19, 20} The Medicare physician’s fee schedule is made available according to current procedural terminology (CPT) codes and the amount paid for a physician’s service is the product of three factors: a nationally standardized RVU for work, practice expense, and malpractice; national geographic practice cost index (GPCI); and an annually adjusted dollar conversion factor (2014 Medicare Conversion Factor = $35.8228).²⁰ The physician fee schedule was calculated using Medicare’s formula: physician payment = [(RVU Work x GPCI Work) + (RVU Practice Expense x GPCI Practice Expense) + (RVU Malpractice x GPCI Malpractice)] x Conversion Factor.²¹ The CPT codes and pricing for incision and drainage specific to superficial SSI: 10180, and for deep/organ space SSI: 74170 (CT abdomen with and without contrast) and 75989 (abscess drainage under CT guidance), and intravenous antibiotics: 96374 + 96376 were utilized for final reimbursement costs. The cost of hospital readmission secondary to surgical site infections was estimated using the Acute Care Hospital Inpatient Prospective Payment System (HIPPS) based on Medicare severity – diagnosis related group (MS-DRG).²² MS-DRGs 857 (postoperative infections with operating room procedure with complications or comorbidities) for superficial SSI and 856 (postoperative infections operating room procedure with major complications or comorbidities) for deep/organ space SSI were applied. A relative weight is assigned to each MS-DRG and inserted in the Medicare’s reimbursement formula: Medicare hospital payment = hospital base rate x relative weight of MS-DRG. The hospital base rate for 2014 of $3 664.21 and relative MS-DRG weight of 2.04 (857) and 4.7874 (856) were used for final calculations.²³

Statistical Analysis

All statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC, USA). Univariate analyses comparing superficial and deep/organ-space SSI with categorical variables were performed using chi-square tests and Wilcoxon rank sum tests for continuous variables. Overall risk of readmission and individual risks secondary to each type of SSI were determined. Multivariate logistic regression models were created for superficial SSI versus no SSI and deep/organ space SSI versus no SSI, as well as for prolonged LOS and unplanned readmission. Patient-specific pre-, intra- and post-operative variables which were considered to be clinically important were adjusted for in the final models regardless of their significance on univariate analyses. Pancreatectomy type was also adjusted for in the models to eliminate potential confounding. P-values of less than 0.05 were considered statistically significant.

RESULTS

A total of 11 219 pancreatectomy procedures were captured in the NSQIP PUF data file. Of these, 8 093 met the study criteria and were analyzed. The overall SSI rate was 17.6% (1 427), 5.2% (422) for superficial and 12.4% (1 005) for deep/organ-space SSI. Median number of days until SSI development was 11 days (IQR: 7 – 17) for superficial, and 12 days (IQR: 8 – 18) for deep/organ-space SSI. A total of 7 088 and 7 671 were included in the multivariate analysis of superficial versus no SSI and deep/organ-space SSI versus no SSI, respectively.

Unadjusted Baseline and Perioperative Characteristics

On univariate analysis, clinical diagnosis type, blood transfusion, open approach, pancreatic duct size, procedure type and prolonged operative time were significantly associated with development of both superficial and deep/organ-space SSI. Patient variables including diabetes mellitus, biliary stent and contaminated or dirty wound class were significant for development of a superficial SSI, while chronic steroid use, history of chronic obstructive pulmonary disease (COPD), body mass index (BMI), gland texture, and concurrent procedures were significant only for deep/organ-space SSI. Female gender and receipt of chemo/radiotherapy were protective of deep/organ-space SSI. The highest significant superficial SSI rates were in patients who had a biliary stent: 10.4% (234), a dirty or contaminated wound: 8.2% (70), and blood transfusion: 8.1% (85), while the lowest rates were found in patients with a minimally invasive approach: 2.6% (42), distal pancreatectomy: 2.8% (64), and those without a biliary stent: 3.9% (188). The highest rates of deep/organ-space SSI were seen in patients with soft gland texture: 18.3% (437), chronic steroid use: 18.1% (39) and concurrent procedure(s): 17.4% (62), while the lowest rates were seen in patients with pancreatitis: 8.6% (35), hard gland texture: 8.9% (171) and minimally invasive approach: 9.0% (153). (Tables 1a and 1b)

Table 1a.

Univariate Analysis of Patient and Pre-Operative Characteristics for Superficial and Deep/Organ-space SSI

	Superficial SSI Rate		Deep/Organ Space SSI Rate
Overall	5.2%		12.4%
Pre-operative Characteristics		** p value		** p value
Age		0.944		0.561
≥ 65 years	5.9%		12.9%
< 65 years	6.0%		13.3%
Sex		0.114		<.0001
Female	5.5%		11.5%
Male	6.4%		14.8%
Race		0.308		0.340
White	6.1%		13.3%
Others/unknown	5.2%		12.2%
ASA Classification		0.306		0.094
ASA III, IV	6.1%		13.5%
ASA I and II	5.5%		12.1%
Smoker		0.282		0.653
Yes	6.6%		13.5%
No	5.8%		13.0%
Chronic Steroid Use		0.723		0.029
Yes	5.4%		18.1%
No	6.0%		13.0%
History of COPD		0.958		0.042
Yes	5.9%		16.8%
No	6.0%		12.9%
Diabetes Mellitus		0.024		0.302
Yes	7.1%		12.4%
No	5.6%		13.3%
Body Mass Index		0.080		<.0001
>30	6.9%		15.9%
25–29.9	5.7%		13.4%
< 25	5.4%		10.3%
Weight Loss > 10 %		0.928		0.589
Yes	6.0%		12.6%
No	5.9%		13.2%
Diagnosis		0.006		0.022
Malignant	6.7%		13.1%
Benign	4.6%		14.7%
Pancreatitis	5.8%		8.6%
Others/unspecified	4.5%		13.4%
Pre-operative Albumin		0.077		0.885
< =3.0	8.1%		12.8%
> 3.0	6.0%		13.0%
***			***
Neoadjuvant
Chemo/radiotherapy		0.230		0.004
Yes	6.8%		10.3%
No	5.8%		13.6%
Preoperative Biliary Stent		<.0001		0.518
Yes	10.4%		13.5%
No /NA	3.9%		12.9%

^***= missing (Superficial: 676; Deep/organ space SSI: 753)

^**= P-value from chi square test

NA = not available

Table 1b.

Univariate Analysis of Intraoperative Characteristics for Superficial and Deep/Organ Space SSI

	Superficial SSI Rate		Deep/Organ Space SSI Rate
Overall	5.2%		12.4%
Intra-operative Characteristics		** p value		** p value
Blood Transfusion		0.002		0.004
Yes	8.1%		15.7%
No	5.6%		12.6%
Approach		<.0001		<.0001
Open	6.9%		14.3%
Minimally Invasive	2.6%		9.0%
Wound Classification		0.003		0.384
Contaminated &
Dirty/Infected	8.2%		12.2%
Clean and Clean-
Contaminated	5.7%		13.2%
Gland Texture		0.060		<.0001
Soft	6.7%		18.3%
Intermediate	6.3%		13.2%
Hard	6.5%		8.9%
Unknown	5.0%		11.6%
Pancreatic Duct Size		0.002		<.0001
<3mm	6.3%		16.9%
3mm	7.0%		11.7%
Unknown	4.8%		12.5%
Vascular resection		0.421		0.973
Yes	6.5%		13.1%
No	5.9%		13.1%
Procedure Type		<.0001		<.0001
Proximal	7.8%		14.9%
Distal	2.8%		10.5%
Others	4.9%		10.0%
Concurrent Procedure		0.750		0.014
Yes	6.4%		17.4%
Null	5.9%		12.9%
Operation Time		<.0001		<.0001
> 305 minutes	7.7%		15.4%
≤ 305 minutes	4.4%		10.9%

^**= P-value from chi square test

Adjusted Baseline and Perioperative Characteristics

After multivariate analysis, having an open approach (OR: 1.63 and 1.60), soft pancreas (OR: 1.39 and 2.22), proximal pancreatectomy (1.50 and 1.42) and each additional hour of operative time (OR: 1.11 and 1.06) were predictive for the development of both superficial and deep/organ-space SSI, respectively (Figures 1a and b). However, patients with preoperative biliary stenting were predisposed to only superficial SSI (OR: 2.21) (Fig. 1a), while those who had BMI between 25 and 29.9 kg/mm² (OR: 1.25) and BMI ≥30 kg/mm² (OR: 1.53) versus BMI <25, pancreatic duct size <3 mm (OR: 1.30) and unknown duct size (OR: 1.45) versus ≥3 mm, intermediate (OR: 1.67) and unknown gland texture (OR: 1.35) versus hard texture, were predisposed to only deep/organ-space SSI (Figure 1b). Females (OR: 0.78), a diagnosis of pancreatitis (OR: 0.62) and having neo-adjuvant chemo/radiotherapy (OR: 0.72) were protective of deep/organ-space SSI (Figure 1b).

Figure 1.

A. Forest plot showing predictors of superficial SSI on multivariate analysis. The following variables were not significant in our final model: gender, race, age, ASA class, smoking status, history of COPD, chronic steroid use, diabetes, BMI, history of weight loss, clinical diagnoses, albumin status, neoadjuvant chemo/radiotherapy, pancreatic duct size, wound class, vascular resection, concurrent procedure and blood transfusion. Unknown albumin levels were treated as missing (676) and were not accounted for in this model.

B. Forest plot showing predictors of deep/organ space SSI on multivariate analysis. The following variables were not significant in our final model: race, age, ASA class, smoking status, history of COPD, diabetes, chronic steroid use, history of weight loss, albumin status, biliary stent, wound class, vascular resection, concurrent procedure and blood transfusion. Unknown albumin levels were treated as missing (753) and were not accounted for in this model.

Secondary Outcomes

Overall median length of stay in our cohort was 7 days (IQR: 5 – 10). The median LOS for patients with superficial SSI (8 days; IQR: 7 – 13) and deep/organ space SSI (11 days; IQR: 7 – 18) was significantly prolonged compared to patients with no SSI (7 days; IQR 5 – 9) on univariate analysis (p<0.0001). On multivariate analyses, superficial and deep/organ-space SSI had significant ORs of 1.74 (95% CI: 1.38 – 2.19) and 1.80 (95% CI: 1.48 – 2.19) for prolonged LOS, respectively, (p <0.0001). Figure 2a shows the forest plot of significant predictors for prolonged LOS.

Figure 2.

A. Forest plot showing significant predictors of prolonged length of hospital stay on multivariate analysis.

The following variables were not significant in our final model: gender, race, BMI, smoking status, history of steroid use, diabetes, BMI, history of weight loss, albumin status, neoadjuvant chemo/radiotherapy, biliary stent, wound class, vascular resection, concurrent procedure, acute renal failure and wound dehiscence Unknown albumin levels were treated as missing (781) and were not accounted for in this model.

B. Forest plot showing significant predictors of unplanned hospital readmission on multivariate analysis.

The following variables were not significant in our final model: age, gender, race, BMI, ASA class, chronic steroid use, diabetes, history of weight loss, albumin status, clinical diagnosis, surgical approach, wound class, type of pancreatectomy, texture of pancreatic gland, pancreatic duct size, vascular resection, concurrent procedure, blood transfusion, reoperation, urinary tract infection, acute renal failure and operation time. Unknown albumin levels were treated as missing (781) and were not accounted for in this model.

The total number of unplanned hospital readmission was 1 330 (16.4%). Of these, the highest primary suspected reasons for readmission were deep/organ space SSI (348; 26.2%), sepsis/septic shock (73; 5.5%) and superficial SSI (44; 3.3%) (Figure 3). There were significantly higher odds of unplanned readmission for superficial SSI: OR of 2.59 (95% CI: 2.00 – 3.35) and deep/organ-space SSI: OR of 6.57 (95% CI: 5.41 – 7.97) (p<0.0001). Figure 2b shows the forest plot of significant predictors of unplanned readmission after adjusting for covariates.

Figure 3.

Pie chart showing the rounding percentage contribution of major contributors to hospital unplanned readmission.

Medicare Costs

In year 2014, the additional estimated costs of readmission per patient secondary to superficial and deep/organ space SSI were $7 661.37 and $18 409.42. (Table 2)

Table 2.

Estimated additional Medicare reimbursement for hospital readmissions secondary to superficial and deep/organ space SSI.

Description	CPT	CPT Definition	Cost of Superficial SSI ($)	Cost of Deep/organ space SSI ($)
Physician Fee	10180
		Incision and drainage, post- operative wound infection	182.0	0
	74170	CT abdomen without and with dye x 2 (Pre- and Post-drainage)	0	560.3
	75989	Abscess drainage under x- ray	0	126.8
Physician Fee (IV antibiotics)	96374 +96376	Intravenous push, single or initial drug + each additional sequential IV push of same drug	0	56.2
Hospital Fee (Relative Weight of MS-DRG 857 x base rate [2.0412 x $3,664.21])			7 479.4	0
Hospital Fee (Relative Weight of MS-DRG 856 x base rate [4.7874 x $3,664.21])			0	17 542.0
IV antibiotics (HCPCS) Piperacillin/Tazobactam ($4.431 for 3.375g every 6 hours x 7 days)			0	124.1
Estimated Cost per patient readmitted			7 661.37	18 409.4

^*Cost estimates were obtained from Centers for Medicare & Medicaid Services (CMS) website ^{19, 20} for 2014 reimbursements. RVU: Relative Value Units; GPCI: Geographic Practice Cost Index; MS-DRG: Medicare Severity- Diagnosis Related Grouping; HCPCS: Healthcare Common Procedure Coding System;

Physician Fee = [(RVU Work x GPCI Work) + (RVU Practice Expense x GPCI Practice Expense) + (RVU Malpractice x GPCI Malpractice)] x Conversion Factor;

Hospital Fee = [Relative Weight of MS-DRG x base rate].

DISCUSSION

Using a multi-institutional database, we found higher rates of deep/organ-space SSI compared with superficial SSI following pancreatectomy. We have also shown distinct predisposing factors for the development of superficial and deep/organ-space SSI, and their varying effects on LOS and hospital readmission. Patients with biliary stenting were found to be predisposed to only superficial SSI, while patients who had a BMI >25kg/m², pancreatic duct size <3mm, soft and intermediate pancreatic gland texture, and proximal pancreatectomy were significantly at risk for deep/organ-space SSI. However, female gender, a diagnosis of pancreatitis and a receipt of neoadjuvant chemo/radiotherapy were protective of deep/organ-space SSI. In addition, after controlling for covariates, both SSI types were found to be independent predictors for prolonged LOS and unplanned readmission. Deep/organ-space SSI had higher odds of prolonged LOS and more than twice the odds of superficial SSI for unplanned readmission. Furthermore, based on conservative estimates of Medicare reimbursements, deep/organ-space SSI, had an estimated cost of $18 409.42 per readmitted patient, which is more than $10 000 higher than the cost of superficial SSI in 2014.

Previous studies evaluating risk factors for pancreatic SSI have traditionally grouped the different types of SSI into one category.^{1, 2, 10, 11, 12, 13} Two separate single institutional retrospective studies, examined risk factors for SSIs by grouping them into incisional (superficial and deep) versus organ-space SSI.^{24, 25} This classification is inconsistent with the current CDC definition for SSIs ⁵ and may inappropriately assume similarities in morbidity and management between the two incisional types. Our SSI classification: 1) superficial and 2) deep/organ-space is congruent with the CDC definition and is also identical to the grouping by Lawson et al.⁴ Studies investigating any SSI found BMI >25, chronic steroid use, biliary stenting, open surgical approach, intra-operative blood transfusion and prolonged operative time to be significant predictors.^{1, 13}

Pancreatic surgery studies have demonstrated higher rates of deep/organ-space SSIs compared with superficial SSI,^{24, 26} in contrast with studies in other major general surgical fields.^{3, 4, 7, 9} Several studies have also attempted to evaluate the cost burden associated with complications following pancreatectomy. Ceppa et al. estimated additional costs of $5 887 and $16 975 for incisional (superficial and deep SSI) and organ-space SSI, respectively.²⁵ Also, a study by Kent et al. estimated an overall additional hospital cost of $15 336 for infectious complications following pancreatectomy, as well as increasing cost gradients according to Clavien-Dindo classification, with additional costs as high as $80 490 in patients with Clavien-Dindo IV.²⁷

Our study extends previous studies on SSIs in pancreatic surgery by using a large cohort to separately evaluate predictors for superficial and deep/organ space SSI. Furthermore, this is the first study to use the recently released targeted pancreatectomy NSQIP module for the investigation of different SSI types, allowing the evaluation of a greater number of potential risk factors. The robust patient mix and hospital variations make our findings more generalizable. We also confirm the findings of studies in colorectal and hepatic surgery who have reported different risk factors for each SSI type,^{3, 4, 7, 8, 9} suggesting different underlying disease processes. Risk factors need to be properly defined to help understand the complex nature of SSIs and reduce the high burden of SSIs in pancreatic surgery. Preoperative biliary stenting in patients undergoing pancreaticoduodenectomy is known to be an important predictor of SSI,^{11, 13, 28, 29} similar to our study. However, after exploring predictors for each SSI type, we found biliary stenting to be predictive of only superficial SSI and not deep/organ-space, contrary to a report by Gavazzi et al.³⁰ Although Teng et al observed a protective trend while comparing 110 patients who received preoperative chemo/radiotherapy, however, no statistical significant association was found. ³¹ Unlike their study, ours was more powered to observe a statistical significant protective association between preoperative chemo/radiotherapy and deep/organ space SSI. This may be attributed to the consequent hardening of pancreatic glands following radiotherapy. By identifying distinct risk factors for different SSIs, surgeons and infectious disease specialists would be able to better understand the burden of SSIs and appropriately direct interventions and resources to decrease their incidence.

There are limitations to our study. The retrospective nature of the data may result in possible misclassification of outcome variables.³² The absence of hospital volume, surgeon volume and geographical information precludes evaluation of local and regional differences in SSI and possible risk adjustment for variations across participating hospitals. Also, we lacked data on certain variables of interest, including microbiology cultures, perioperative antibiotic use and sensitivity, anastomotic techniques, additional therapies like somatostain analogues which limited conclusions that could be drawn. Our estimated costs for each readmitted patient secondary to SSIs are conservative and substantially underestimate the true economic burden of deep/organ space SSI, due to lack of data on post-operative evaluations, such as blood specimens, microbiological and radiological work-ups, transportation to facility, rehabilitative facility care and home nursing care. Also, the cost estimates may not be generalizable since only the CMS recommendations were used, which may be different from the reimbursements of private insurance providers.

Despite these limitations, our study is the first multi-institutional study to separately evaluate predictors of SSI types and their impact on hospital cost. The targeted pancreatic NSQIP module is a validated database with high data quality measures put in place to ensure accuracy of data and includes more variables than the general NSQIP.^{15, 33} Also, it is specifically designed for 30 day outcomes and, therefore, appropriate for evaluation of SSIs. A large sample size makes our data more robust and provides enough power to detect small differences. We were able to identify different risk factors for the development of each type of SSI and their impact on health care costs. Even though conservative cost estimates were derived, additional costs secondary to SSIs were high, with deep/organ space SSI having a more profound impact on Medicare reimbursement costs.

Targeting SSI types differently could improve outcomes of national quality improvement programs such as the surgical care improvement program (SCIP) and may help provide separate perioperative surgical care bundles, which have proven to be effective in reducing SSIs.^{34, 35} Additional studies will be required to assess microorganism susceptibility and antibiotic regimens that will be effective in reducing each type of SSI. Based on our findings, quality improvement strategies and hospital resources should be directed at preventing deep/organ space SSI because of their significant impact on post-operative outcomes and cost of care.

Table 1c.

Univariate Analysis of Postoperative Characteristics for Superficial and Deep/Organ Space SSI

	Superficial SSI Rate		Deep/Organ Space SSI Rate
Overall	5.2%		12.4%
Post-operative Characteristics		** p value		** p value
Fistula		<.0001		<.0001
Yes	10.8%		43.9%
No/unknown	5.3%		6.6%
Length of Stay (days)		<.0001		<.0001
7.0	9.3%		21.0%
7.0	3.8%		7.4%
Unplanned Readmission		<.0001		<.0001
Yes	12.5%		42.6%
No	5.1%		7.5%
Mortality		*0.627		<.0001
Yes	4.1%		27.6%
No	6.0%		12.9%

^**= P-value from chi square test

^*= P-value from fisher’s exact test